Attempt to fix a hang in the GPU process while creating a D3D9 device.

Attempt to fix a hang in the GPU process while creating a D3D9 device.

Mimicing some of the checks in the firefox sources to avoid creating a device
if we find that the driver does not support H/W format conversion and to
use D3DCREATE_MIXED_VERTEXPROCESSING instead of D3DCREATE_HARDWARE_VERTEXPROCESSING. As per docs it means
that the driver falls back to software vertex processing
which is still better than doing the whole decoding on the CPU.

Additionally I removed the D3DCREATE_DISABLE_PSGP_THREADING flag which restricts
computation to the main thread in case the driver fallsback to software vertex
processing which could affect performance.

The expectation with this patch is we may be able to avoid device creation for
a bunch of buggy drivers.

BUG=602389
Committed: https://crrev.com/30968a7ee88a9efbd92de708f7d7564ab3d29451
Cr-Commit-Position: refs/heads/master@{#387738}

[ananta, 2016-04-15 02:13:18]

Description was changed from

==========
Attempt to fix a hang in the GPU process while creating a D3D9 device.

Mimicing some of the checks in the firefox sources to avoid creating a device
if we find that the driver does not support H/W format conversion and to
use D3DCREATE_MIXED_VERTEXPROCESSING instead of
D3DCREATE_HARDWARE_VERTEXPROCESSING.

Additionally I removed the D3DCREATE_DISABLE_PSGP_THREADING flag which restricts
computation to the main thread in case the driver fallsback to software vertex
processing which could affect performance.

The expectation with this patch is we may be able to avoid device creation for
a bunch of buggy drivers.

BUG=602389
==========

to

==========
Attempt to fix a hang in the GPU process while creating a D3D9 device.

Mimicing some of the checks in the firefox sources to avoid creating a device
if we find that the driver does not support H/W format conversion and to
use D3DCREATE_MIXED_VERTEXPROCESSING instead of
D3DCREATE_HARDWARE_VERTEXPROCESSING. As per docs it means
that the driver falls back to software vertex processing 
which is still better than doing the whole decoding on the CPU.

Additionally I removed the D3DCREATE_DISABLE_PSGP_THREADING flag which restricts
computation to the main thread in case the driver fallsback to software vertex
processing which could affect performance.

The expectation with this patch is we may be able to avoid device creation for
a bunch of buggy drivers.

BUG=602389
==========

[ananta, 2016-04-15 02:13:50]

ananta@chromium.org changed reviewers:
+ sandersd@chromium.org

[sandersd (OOO until July 31), 2016-04-15 23:03:03]

lgtm

[ananta, 2016-04-15 23:03:49]

The CQ bit was checked by ananta@chromium.org

[commit-bot: I haz the power, 2016-04-15 23:04:29]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1884133004/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1884133004/1

[commit-bot: I haz the power, 2016-04-15 23:10:49]

Description was changed from

==========
Attempt to fix a hang in the GPU process while creating a D3D9 device.

Mimicing some of the checks in the firefox sources to avoid creating a device
if we find that the driver does not support H/W format conversion and to
use D3DCREATE_MIXED_VERTEXPROCESSING instead of
D3DCREATE_HARDWARE_VERTEXPROCESSING. As per docs it means
that the driver falls back to software vertex processing 
which is still better than doing the whole decoding on the CPU.

Additionally I removed the D3DCREATE_DISABLE_PSGP_THREADING flag which restricts
computation to the main thread in case the driver fallsback to software vertex
processing which could affect performance.

The expectation with this patch is we may be able to avoid device creation for
a bunch of buggy drivers.

BUG=602389
==========

to

==========
Attempt to fix a hang in the GPU process while creating a D3D9 device.

Mimicing some of the checks in the firefox sources to avoid creating a device
if we find that the driver does not support H/W format conversion and to
use D3DCREATE_MIXED_VERTEXPROCESSING instead of
D3DCREATE_HARDWARE_VERTEXPROCESSING. As per docs it means
that the driver falls back to software vertex processing 
which is still better than doing the whole decoding on the CPU.

Additionally I removed the D3DCREATE_DISABLE_PSGP_THREADING flag which restricts
computation to the main thread in case the driver fallsback to software vertex
processing which could affect performance.

The expectation with this patch is we may be able to avoid device creation for
a bunch of buggy drivers.

BUG=602389
==========

[commit-bot: I haz the power, 2016-04-15 23:10:49]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2016-04-15 23:12:12]

Description was changed from

==========
Attempt to fix a hang in the GPU process while creating a D3D9 device.

Mimicing some of the checks in the firefox sources to avoid creating a device
if we find that the driver does not support H/W format conversion and to
use D3DCREATE_MIXED_VERTEXPROCESSING instead of
D3DCREATE_HARDWARE_VERTEXPROCESSING. As per docs it means
that the driver falls back to software vertex processing 
which is still better than doing the whole decoding on the CPU.

Additionally I removed the D3DCREATE_DISABLE_PSGP_THREADING flag which restricts
computation to the main thread in case the driver fallsback to software vertex
processing which could affect performance.

The expectation with this patch is we may be able to avoid device creation for
a bunch of buggy drivers.

BUG=602389
==========

to

==========
Attempt to fix a hang in the GPU process while creating a D3D9 device.

Mimicing some of the checks in the firefox sources to avoid creating a device
if we find that the driver does not support H/W format conversion and to
use D3DCREATE_MIXED_VERTEXPROCESSING instead of
D3DCREATE_HARDWARE_VERTEXPROCESSING. As per docs it means
that the driver falls back to software vertex processing
which is still better than doing the whole decoding on the CPU.

Additionally I removed the D3DCREATE_DISABLE_PSGP_THREADING flag which restricts
computation to the main thread in case the driver fallsback to software vertex
processing which could affect performance.

The expectation with this patch is we may be able to avoid device creation for
a bunch of buggy drivers.

BUG=602389

Committed: https://crrev.com/30968a7ee88a9efbd92de708f7d7564ab3d29451
Cr-Commit-Position: refs/heads/master@{#387738}
==========

[commit-bot: I haz the power, 2016-04-15 23:12:12]

Patchset 1 (id:??) landed as
https://crrev.com/30968a7ee88a9efbd92de708f7d7564ab3d29451
Cr-Commit-Position: refs/heads/master@{#387738}